java - Trying to make simple PDF document with Apache poi

Question

I see the internet is riddled with people complaining about apache's pdf products, but I cannot find my particular usecase here. I am trying to do a simple Hello World with apache poi. Right now my code is as follows:

public ByteArrayOutputStream export() throws IOException {
    //Blank Document
    XWPFDocument document = new XWPFDocument();

    //Write the Document in file system
    ByteArrayOutputStream out = new ByteArrayOutputStream();;

    //create table
    XWPFTable table = document.createTable();
    XWPFStyles styles = document.createStyles();
    styles.setSpellingLanguage("English");
    //create first row
    XWPFTableRow tableRowOne = table.getRow(0);
    tableRowOne.getCell(0).setText("col one, row one");
    tableRowOne.addNewTableCell().setText("col two, row one");
    tableRowOne.addNewTableCell().setText("col three, row one");

    //create second row
    XWPFTableRow tableRowTwo = table.createRow();
    tableRowTwo.getCell(0).setText("col one, row two");
    tableRowTwo.getCell(1).setText("col two, row two");
    tableRowTwo.getCell(2).setText("col three, row two");

    //create third row
    XWPFTableRow tableRowThree = table.createRow();
    tableRowThree.getCell(0).setText("col one, row three");
    tableRowThree.getCell(1).setText("col two, row three");
    tableRowThree.getCell(2).setText("col three, row three");

    PdfOptions options = PdfOptions.create();
    PdfConverter.getInstance().convert(document, out, options);
    out.close();
    return out;
}

and the code that calls this is:

    public ResponseEntity<Resource> convertToPDFPost(@ApiParam(value = "DTOs passed from the FE" ,required=true )  @Valid @RequestBody ExportEnvelopeDTO exportDtos) {

        if (exportDtos.getProdExportDTOs() != null) {
            try {
                FileOutputStream out = new FileOutputStream("/Users/kornhaus/Desktop/test.pdf");
                out.write(exporter.export().toByteArray());
                out.close();
            } catch (IOException e) {
                e.printStackTrace();
            }
            return new ResponseEntity<Resource>(responseFile, responseHeaders, HttpStatus.OK);
        }

        return new ResponseEntity<Resource>(HttpStatus.INTERNAL_SERVER_ERROR);
    }

}

On this line here: out.write(exporter.export().toByteArray()); the code throws an exception:

org.apache.poi.xwpf.converter.core.XWPFConverterException: java.io.IOException: Unable to parse xml bean

I have no clue what's causing this, where to even look for this kind of documentation. I have been coding a decade plus and never had such difficulty with what should be a simple Java library. Any help would be great.

Answer 1

The main problem with this is that those PdfOptions and PdfConverter are not part of the apache poi project. They are developed by opensagres and first versions were badly named org.apache.poi.xwpf.converter.pdf.PdfOptions and org.apache.poi.xwpf.converter.pdf.PdfConverter. Those old classes were not updated since 2014 and needs version 3.9 of apache poi to be used.

But the same developers provide fr.opensagres.poi.xwpf.converter.pdf, which is much more current and works using the latest stable release apache poi 3.17. So we should using this.

But since even those newer PdfOptions and PdfConverter are not part of the apache poi project, apache poi will not testing those with their releases. And so the default *.docx documents created by apache poi lacks some content which PdfConverter needs.

1. There must be a styles document, even if it is empty.

2. There must be section properties for the page having at least the page size set.

3. Tables must have a table grid set.

To fulfilling this we must add some code additionally in our program. Unfortunately this then needs the full jar of all of the schemas ooxml-schemas-1.3.jar as mentioned in Faq-N10025.

And because we need changing the underlaying low level objects, the document must be written so underlaying objects will be committed. Else the XWPFDocument which we hand over the PdfConverter will be incomplete.

Example:

import java.io.*;
import java.math.BigInteger;

//needed jars: fr.opensagres.poi.xwpf.converter.core-2.0.1.jar, 
//             fr.opensagres.poi.xwpf.converter.pdf-2.0.1.jar,
//             fr.opensagres.xdocreport.itext.extension-2.0.1.jar,
//             itext-2.1.7.jar                                  
import fr.opensagres.poi.xwpf.converter.pdf.PdfOptions;
import fr.opensagres.poi.xwpf.converter.pdf.PdfConverter;

//needed jars: apache poi and it's dependencies
//             and additionally: ooxml-schemas-1.3.jar 
import org.apache.poi.xwpf.usermodel.*;
import org.apache.poi.util.Units;
import org.openxmlformats.schemas.wordprocessingml.x2006.main.*;

public class XWPFToPDFConverterSampleMin {

 public static void main(String[] args) throws Exception {

  XWPFDocument document = new XWPFDocument();

  // there must be a styles document, even if it is empty
  XWPFStyles styles = document.createStyles();

  // there must be section properties for the page having at least the page size set
  CTSectPr sectPr = document.getDocument().getBody().addNewSectPr();
  CTPageSz pageSz = sectPr.addNewPgSz();
  pageSz.setW(BigInteger.valueOf(12240)); //12240 Twips = 12240/20 = 612 pt = 612/72 = 8.5"
  pageSz.setH(BigInteger.valueOf(15840)); //15840 Twips = 15840/20 = 792 pt = 792/72 = 11"

  // filling the body
  XWPFParagraph paragraph = document.createParagraph();

  //create table
  XWPFTable table = document.createTable();

  //create first row
  XWPFTableRow tableRowOne = table.getRow(0);
  tableRowOne.getCell(0).setText("col one, row one");
  tableRowOne.addNewTableCell().setText("col two, row one");
  tableRowOne.addNewTableCell().setText("col three, row one");

  //create CTTblGrid for this table with widths of the 3 columns. 
  //necessary for Libreoffice/Openoffice and PdfConverter to accept the column widths.
  //values are in unit twentieths of a point (1/1440 of an inch)
  //first column = 2 inches width
  table.getCTTbl().addNewTblGrid().addNewGridCol().setW(BigInteger.valueOf(2*1440));
  //other columns (2 in this case) also each 2 inches width
  for (int col = 1 ; col < 3; col++) {
   table.getCTTbl().getTblGrid().addNewGridCol().setW(BigInteger.valueOf(2*1440));
  }

  //create second row
  XWPFTableRow tableRowTwo = table.createRow();
  tableRowTwo.getCell(0).setText("col one, row two");
  tableRowTwo.getCell(1).setText("col two, row two");
  tableRowTwo.getCell(2).setText("col three, row two");

  //create third row
  XWPFTableRow tableRowThree = table.createRow();
  tableRowThree.getCell(0).setText("col one, row three");
  tableRowThree.getCell(1).setText("col two, row three");
  tableRowThree.getCell(2).setText("col three, row three");

  paragraph = document.createParagraph();

  //trying picture
  XWPFRun run = paragraph.createRun();
  run.setText("The picture in line: ");
  InputStream in = new FileInputStream("samplePict.jpeg");
  run.addPicture(in, Document.PICTURE_TYPE_JPEG, "samplePict.jpeg", Units.toEMU(100), Units.toEMU(30));
  in.close();  
  run.setText(" text after the picture.");

  paragraph = document.createParagraph();

  //document must be written so underlaaying objects will be committed
  ByteArrayOutputStream out = new ByteArrayOutputStream();
  document.write(out);
  document.close();

  document = new XWPFDocument(new ByteArrayInputStream(out.toByteArray()));
  PdfOptions options = PdfOptions.create();
  PdfConverter converter = (PdfConverter)PdfConverter.getInstance();
  converter.convert(document, new FileOutputStream("XWPFToPDFConverterSampleMin.pdf"), options);

  document.close();

 }
}

---

Using XDocReport

Another way would be using the newest version of opensagres/xdocreport as described in Converter only with ConverterRegistry:

import java.io.*;
import java.math.BigInteger;

//needed jars: xdocreport-2.0.1.jar, 
//             odfdom-java-0.8.7.jar,
//             itext-2.1.7.jar  
import fr.opensagres.xdocreport.converter.Options;
import fr.opensagres.xdocreport.converter.IConverter;
import fr.opensagres.xdocreport.converter.ConverterRegistry;
import fr.opensagres.xdocreport.converter.ConverterTypeTo;
import fr.opensagres.xdocreport.core.document.DocumentKind;

//needed jars: apache poi and it's dependencies
//             and additionally: ooxml-schemas-1.3.jar 
import org.apache.poi.xwpf.usermodel.*;
import org.apache.poi.util.Units;
import org.openxmlformats.schemas.wordprocessingml.x2006.main.*;

public class XWPFToPDFXDocReport {

 public static void main(String[] args) throws Exception {

  XWPFDocument document = new XWPFDocument();

  // there must be a styles document, even if it is empty
  XWPFStyles styles = document.createStyles();

  // there must be section properties for the page having at least the page size set
  CTSectPr sectPr = document.getDocument().getBody().addNewSectPr();
  CTPageSz pageSz = sectPr.addNewPgSz();
  pageSz.setW(BigInteger.valueOf(12240)); //12240 Twips = 12240/20 = 612 pt = 612/72 = 8.5"
  pageSz.setH(BigInteger.valueOf(15840)); //15840 Twips = 15840/20 = 792 pt = 792/72 = 11"

  // filling the body
  XWPFParagraph paragraph = document.createParagraph();

  //create table
  XWPFTable table = document.createTable();

  //create first row
  XWPFTableRow tableRowOne = table.getRow(0);
  tableRowOne.getCell(0).setText("col one, row one");
  tableRowOne.addNewTableCell().setText("col two, row one");
  tableRowOne.addNewTableCell().setText("col three, row one");

  //create CTTblGrid for this table with widths of the 3 columns. 
  //necessary for Libreoffice/Openoffice and PdfConverter to accept the column widths.
  //values are in unit twentieths of a point (1/1440 of an inch)
  //first column = 2 inches width
  table.getCTTbl().addNewTblGrid().addNewGridCol().setW(BigInteger.valueOf(2*1440));
  //other columns (2 in this case) also each 2 inches width
  for (int col = 1 ; col < 3; col++) {
   table.getCTTbl().getTblGrid().addNewGridCol().setW(BigInteger.valueOf(2*1440));
  }

  //create second row
  XWPFTableRow tableRowTwo = table.createRow();
  tableRowTwo.getCell(0).setText("col one, row two");
  tableRowTwo.getCell(1).setText("col two, row two");
  tableRowTwo.getCell(2).setText("col three, row two");

  //create third row
  XWPFTableRow tableRowThree = table.createRow();
  tableRowThree.getCell(0).setText("col one, row three");
  tableRowThree.getCell(1).setText("col two, row three");
  tableRowThree.getCell(2).setText("col three, row three");

  paragraph = document.createParagraph();

  //trying picture
  XWPFRun run = paragraph.createRun();
  run.setText("The picture in line: ");
  InputStream in = new FileInputStream("samplePict.jpeg");
  run.addPicture(in, Document.PICTURE_TYPE_JPEG, "samplePict.jpeg", Units.toEMU(100), Units.toEMU(30));
  in.close();  
  run.setText(" text after the picture.");

  paragraph = document.createParagraph();

  //document must be written so underlaaying objects will be committed
  ByteArrayOutputStream out = new ByteArrayOutputStream();
  document.write(out);
  document.close();

  // 1) Create options DOCX 2 PDF to select well converter form the registry
  Options options = Options.getFrom(DocumentKind.DOCX).to(ConverterTypeTo.PDF);

  // 2) Get the converter from the registry
  IConverter converter = ConverterRegistry.getRegistry().getConverter(options);

  // 3) Convert DOCX 2 PDF
  InputStream docxin= new ByteArrayInputStream(out.toByteArray());
  OutputStream pdfout = new FileOutputStream(new File("XWPFToPDFXDocReport.pdf"));
  converter.convert(docxin, pdfout, options);

  docxin.close();       
  pdfout.close();       
 
 }
}

---

October 2018: This code works using apache poi 3.17. It cannot work using apache poi 4.0.0 due to changings in apache poi which were not taken in account until now in fr.opensagres.poi.xwpf.converter as well as in fr.opensagres.xdocreport.converter.

---

February 2019: Works for me now using the newest apache poi version 4.0.1 and the newest version 2.0.2 of fr.opensagres.poi.xwpf.converter.core and consorts.

---

June 2021: Works using apache poi version 4.1.2 and the newest version 2.0.2 of fr.opensagres.poi.xwpf.converter.core and consorts. Cannot work using apache poi version 5.0.0 because XDocReport needs ooxml-schemas which apache poi 5 does not support anymore.